Sample policy language
1. Purpose
We use approved AI tools to support drafting, research, organization, and routine administrative work. AI output is a starting point. An accountable person remains responsible for accuracy, fairness, confidentiality, and the final action.
2. Approved use
Employees may use company-approved tools for low-risk work that does not expose confidential, personal, regulated, or client-protected information. Users must check source material, correct errors, and identify AI-assisted work when required by a client, contract, or company process.
3. Prohibited use
Do not enter employee, applicant, medical, accommodation, complaint, investigation, payroll, financial, customer, credential, or trade-secret information into an unapproved AI tool. Do not ask AI to make hiring, pay, discipline, termination, safety, medical, or legal decisions.
4. Human review
The process owner must review the output against approved sources before it is used. High-risk or unclear situations must be escalated to the appropriate HR, legal, IT-security, safety, privacy, or other qualified professional.
5. Records and reporting
Save required business records in approved systems—not only in an AI chat. Promptly report accidental disclosure, suspicious output, policy violations, or a safety or employment concern through the company's normal incident process.
Implementation checklist
List which tools may be used, for what purpose, and who approves changes.
Give examples from your business so employees know what must never be pasted into an unapproved tool.
Practice a safe draft, an unsafe request, and the correct escalation path.
Revisit tools, contracts, security controls, incidents, and legal requirements at least annually and when systems change.